Effect of a Best Practice Alert on Birth-Cohort Screening for Hepatitis C Virus

Abstract

Introduction: We assessed the influence of a best practice alert (BPA) embedded within the electronic medical record on improving hepatitis C virus (HCV) birth-cohort screening by primary care physicians (PCPs).

Methods: Screening by 155 PCPs was monitored during 2 consecutive 9-month periods before and after implementation of the BPA. All tests were reviewed to differentiate true screening from other testing indications.

Results: Of 155 PCPs, 131 placed screening orders before and after BPA. Twenty-two PCPs started testing after BPA (P = 0.02). The number of tests placed and screening rates per PCP increased from 16 to 84 and from 3.3% to 13.2%, respectively (P < 0.0001). Before BPA, most PCPs rarely ordered screening HCV tests, whereas a small group of physicians generated most tests, indicative of an underlying power-law distribution. After the BPA, a new group of high-performing PCPs emerged, whose screening patterns were again characterized by a power-law distribution. However, pre-BPA test rates of individual PCPs were not predictive of their post-BPA rates. Overall, the introduction of the BPA narrowed the gap between low- and high-performing testers, indicating that modest increases in testing by a large number of low-performing PCPs could drive substantial improvement in program implementation.

Discussion: HCV birth-cohort screening by PCPs was shaped by an underlying power-law distribution. This distribution was preserved after the implementation of a BPA, although pre-BPA test rates were not predictive of post-BPA rates. Increases in test rates by high- and low-performing PCPs both contributed to the overall success of the BPA.

Figure 1.

HCV testing follows a power-law distribution. The x axes display the number of patients for whom HCV test orders were placed by each individual PCP, in bins of 5 (a) or 10 (b) patients. The y axis indicates the number of PCPs placing orders for each bin. a/b and c/d represent pre-BPA and post-BPA analyses, respectively. Dual logarithmic transformation of the data in a and c revealed significant linear correlations, consistent with a power-law distribution (b and d). BPA, best practice alert; HCV, hepatitis C virus; PCP, primary care physician.

Figure 2.

Pre-BPA screening performance does not predict post-BPA performance. The x axes display the absolute number (upper panel) or the percentage (lower panel) of screening-eligible patients, for whom test orders were placed by individual PCPs before the BPA. The y axes display the corresponding post-BPA number or percentage. Linear regression analyses showed no significant correlations. BPA, best practice alert; PCP, primary care physician.

Figure 3.

The relative contributions of high- and low-performing PCPs to the overall screening effort before BPA and after BPA. Before BPA, the top 20% and bottom 80% of PCPs generated 72% and 28% of all test orders, respectively. The corresponding results after BPA were 54% and 46%, respectively. BPA, best practice alert; PCP, primary care physician.